In this project we will develop a computational approach to model membrane proteins for which a limited number of experimental restraints are available but for which the experimental structure is difficult to obtain. We will utilize our recently developed fragment library of supersecondary structure elements (Smotifs) that exhaustively classifies all known building blocks of proteins. Recently we have shown that this library of Smotifs saturated almost 10 years ago, and that new folds seem to be a novel combination of existing Smotifs. Therefore we hypothesize that all protein folds should be possible to build from this library. In order to model membrane proteins we can calculate hypothetical chemical shift values for all our Smotifs, while chemical shift values for a protein of interest can usually be quickly and easily obtained and assigned from initial NMR experiments. This proposal is concerned with developing algorithms that can match experimentally observed and theoretically calculated chemical shift patterns of Smotifs and therefore identify a subset of Smotif conformations that form a protein. The second part of the proposal is concerned of setting up an optimization approach (a sampling algorithm along the degrees of freedom of Smotif combinations and a scoring function) that will rapidly assemble overlapping Smotifs into compact folds using additional experimental restraints obtained from NMR dipolar coupling data. In later years of the project we will apply our technique on specific proteins for which chemical shift and dipolar coupling data were obtained and subsequently verify our computational models with spin labeling experiments. The technologies developed in this application will provide the foundation required for efficient modeling of membrane proteins for which a very limited number of experimental structures are available in the PDB. Meanwhile membrane proteins constitute the majority of targets of currently known drugs. Our effort is focused on increasing the rate of discovering membrane protein structures and therefore will lay a foundation for more effective rational drug design. PUBLIC HEALTH RELEVANCE: The majority of currently known drugs target membrane proteins, of which only about 0.5% have been structurally characterized. In this proposal we will develop a fragment assembly modeling approach that takes advantage of NMR chemical shift data and our recently developed supersecondary structure library. Our effort is concerned with increasing the rate of discovering membrane protein structures and will lay a foundation for effective rational drug design for this important class of proteins.